From 6276d667177d915aff5c5b86e895a9b38b7ea2ae Mon Sep 17 00:00:00 2001
From: Jean-Sebastien Pedron <jean-sebastien@rabbitmq.com>
Date: Thu, 27 Nov 2014 12:13:09 +0100
Subject: Check cluster consistency before joining a cluster

Before this change, the node would join the cluster and the consistency
would be only checked during application restart (rabbitmqctl
start_app). This could lead to crash during the join due to Erlang,
Mnesia or RabbitMQ incompatibilities.

Now, the join_cluster command reports the problem. Here's an example:
    $ rabbitmqctl join_cluster <remote_node>

    Clustering node <local_node> with <remote_node> ...
    Error: {inconsistent_cluster,
            "OTP version mismatch: local node is R16B03-1, remote node 17.3"}
---
 src/rabbit_mnesia.erl | 42 +++++++++++++++++++++++++++---------------
 1 file changed, 27 insertions(+), 15 deletions(-)

diff --git a/src/rabbit_mnesia.erl b/src/rabbit_mnesia.erl
index fa51dd70..b55d53b6 100644
--- a/src/rabbit_mnesia.erl
+++ b/src/rabbit_mnesia.erl
@@ -174,18 +174,25 @@ join_cluster(DiscoveryNode, NodeType) ->
     {ClusterNodes, _, _} = discover_cluster([DiscoveryNode]),
     case me_in_nodes(ClusterNodes) of
         false ->
-            %% reset the node. this simplifies things and it will be needed in
-            %% this case - we're joining a new cluster with new nodes which
-            %% are not in synch with the current node. I also lifts the burden
-            %% of reseting the node from the user.
-            reset_gracefully(),
-
-            %% Join the cluster
-            rabbit_log:info("Clustering with ~p as ~p node~n",
-                            [ClusterNodes, NodeType]),
-            ok = init_db_with_mnesia(ClusterNodes, NodeType, true, true),
-            rabbit_node_monitor:notify_joined_cluster(),
-            ok;
+            case check_cluster_consistency(DiscoveryNode, false) of
+                {ok, _} ->
+                    %% reset the node. this simplifies things and it
+                    %% will be needed in this case - we're joining a new
+                    %% cluster with new nodes which are not in synch
+                    %% with the current node. It also lifts the burden
+                    %% of resetting the node from the user.
+                    reset_gracefully(),
+
+                    %% Join the cluster
+                    rabbit_log:info("Clustering with ~p as ~p node~n",
+                                    [ClusterNodes, NodeType]),
+                    ok = init_db_with_mnesia(ClusterNodes, NodeType,
+                                             true, true),
+                    rabbit_node_monitor:notify_joined_cluster(),
+                    ok;
+                {error, Reason} ->
+                    {error, Reason}
+            end;
         true ->
             rabbit_log:info("Already member of cluster: ~p~n", [ClusterNodes]),
             {ok, already_member}
@@ -545,7 +552,7 @@ maybe_force_load() ->
 check_cluster_consistency() ->
     %% We want to find 0 or 1 consistent nodes.
     case lists:foldl(
-           fun (Node,  {error, _})    -> check_cluster_consistency(Node);
+           fun (Node,  {error, _})    -> check_cluster_consistency(Node, true);
                (_Node, {ok, Status})  -> {ok, Status}
            end, {error, not_found}, nodes_excl_me(cluster_nodes(all)))
     of
@@ -575,17 +582,22 @@ check_cluster_consistency() ->
             throw(E)
     end.
 
-check_cluster_consistency(Node) ->
+check_cluster_consistency(Node, AlreadyFormed) ->
     case rpc:call(Node, rabbit_mnesia, node_info, []) of
         {badrpc, _Reason} ->
             {error, not_found};
         {_OTP, _Rabbit, {error, _}} ->
             {error, not_found};
-        {OTP, Rabbit, {ok, Status}} ->
+        {OTP, Rabbit, {ok, Status}} when AlreadyFormed ->
             case check_consistency(OTP, Rabbit, Node, Status) of
                 {error, _} = E -> E;
                 {ok, Res}      -> {ok, Res}
             end;
+        {OTP, Rabbit, {ok, Status}} ->
+            case check_consistency(OTP, Rabbit) of
+                {error, _} = E -> E;
+                ok             -> {ok, Status}
+            end;
         {_OTP, Rabbit, _Hash, _Status} ->
             %% delegate hash checking implies version mismatch
             version_error("Rabbit", rabbit_misc:version(), Rabbit)
-- 
cgit v1.2.1


From f6cfb1a6195aca0af66652b918160dcce02ce34b Mon Sep 17 00:00:00 2001
From: Simon MacMullen <simon@rabbitmq.com>
Date: Fri, 28 Nov 2014 10:41:59 +0000
Subject: Rename this to indicate what it does, not the current circumstances
 in which it is called.

---
 src/rabbit_mnesia.erl | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/src/rabbit_mnesia.erl b/src/rabbit_mnesia.erl
index b55d53b6..80fbcd68 100644
--- a/src/rabbit_mnesia.erl
+++ b/src/rabbit_mnesia.erl
@@ -582,13 +582,13 @@ check_cluster_consistency() ->
             throw(E)
     end.
 
-check_cluster_consistency(Node, AlreadyFormed) ->
+check_cluster_consistency(Node, CheckNodesConsistency) ->
     case rpc:call(Node, rabbit_mnesia, node_info, []) of
         {badrpc, _Reason} ->
             {error, not_found};
         {_OTP, _Rabbit, {error, _}} ->
             {error, not_found};
-        {OTP, Rabbit, {ok, Status}} when AlreadyFormed ->
+        {OTP, Rabbit, {ok, Status}} when CheckNodesConsistency ->
             case check_consistency(OTP, Rabbit, Node, Status) of
                 {error, _} = E -> E;
                 {ok, Res}      -> {ok, Res}
-- 
cgit v1.2.1